AB TEST

本报告详细分析了游戏内购活动的表现,通过SQL查询,对比了不同时间段内玩家的消费行为,包括支付总额、购买次数及平均消费额,为游戏运营提供数据支持。

五号桶到现在11天

with rank as(
SELECT group_id, SUM(pay_dollar) as pay_dollars, count(*) as count FROM
(select mod(player_id, 33) as group_id,  pay_amount*exchange as pay_dollar FROM mafia1.offer_purchase
WHERE timestamp >='2019-10-19 03:10:00' and timestamp <='2019-10-30 06:10:00' and player_id not in (select player_id FROM mafia1.create_player where uid in (select uid from mafia1.internal_user)))
# and player_id in (select player_id FROM mafia1.create_player where timestamp >='2019-07-01'))
#where pay_dollar >=90
group by group_id
order by pay_dollars)

select group_id, pay_dollars/total_dollars as ratio, count, pay_dollars, pay_dollars/count as mean FROM rank, (select SUM(pay_dollars) as total_dollars FROM rank)
Order by pay_dollars 

五号桶上线前11天

with rank as(
SELECT group_id, SUM(pay_dollar) as pay_dollars, count(*) as count FROM
(select mod(player_id, 33) as group_id,  pay_amount*exchange as pay_dollar FROM mafia1.offer_purchase
WHERE timestamp >='2019-10-06 03:10:00' and timestamp <='2019-10-17 01:30:00' and player_id not in (select player_id FROM mafia1.create_player where uid in (select uid from mafia1.internal_user)))
# and player_id in (select player_id FROM mafia1.create_player where timestamp >='2019-07-01'))
#where pay_dollar >=90
group by group_id
order by pay_dollars)

select group_id, pay_dollars/total_dollars as ratio, count, pay_dollars, pay_dollars/count as mean FROM rank, (select SUM(pay_dollars) as total_dollars FROM rank)
Order by pay_dollars 

29号桶到现在7天

with rank as(
SELECT group_id, SUM(pay_dollar) as pay_dollars, count(*) as count FROM
(select mod(player_id, 33) as group_id,  pay_amount*exchange as pay_dollar FROM mafia1.offer_purchase
WHERE timestamp >='2019-10-24 06:10:00' and timestamp <='2019-10-30 06:10:00' and player_id not in (select player_id FROM mafia1.create_player where uid in (select uid from mafia1.internal_user)))
# and player_id in (select player_id FROM mafia1.create_player where timestamp >='2019-07-01'))
#where pay_dollar >=90
group by group_id
order by pay_dollars)

select group_id, pay_dollars/total_dollars as ratio, count, pay_dollars, pay_dollars/count as mean FROM rank, (select SUM(pay_dollars) as total_dollars FROM rank)
Order by pay_dollars 

29号桶上线前7天

with rank as(
SELECT group_id, SUM(pay_dollar) as pay_dollars, count(*) as count FROM
(select mod(player_id, 33) as group_id,  pay_amount*exchange as pay_dollar FROM mafia1.offer_purchase
WHERE timestamp >='2019-10-18 06:10:00' and timestamp <='2019-10-24 06:10:00' and player_id not in (select player_id FROM mafia1.create_player where uid in (select uid from mafia1.internal_user)))
# and player_id in (select player_id FROM mafia1.create_player where timestamp >='2019-07-01'))
#where pay_dollar >=90
group by group_id
order by pay_dollars)

select group_id, pay_dollars/total_dollars as ratio, count, pay_dollars, pay_dollars/count as mean FROM rank, (select SUM(pay_dollars) as total_dollars FROM rank)
Order by pay_dollars 

09-10
ABtest的底层逻辑是验证用户对两个版本的认知,其原理基于控制唯一变量和设置必要的试验周期,通过对比不同版本来评估效果。但由于需要控制唯一变量和必要的试验周期,能进行的abtest次数会受到局限,例如仅有10%的流量用于实验,多个业务都想用,就需要分层协调,包括确定谁先用、谁灰度、谁等候等,效率较低,此时可进行分层操作,但有一定前提条件 [^2][^3]。 在应用场景方面,不同场景下ABtest功能的实现及构建方式不同,没有相同、完全一样的服务。对于垂类、用户关联性强的产品/部门,可以把AbTest放置开机阶段,开机往往具有热加载及倒计时逻辑,可以融纳较多的服务;对于Feed类、用户关联性弱的产品/部门,可以考虑把AbTest放置在前置或后置服务,这样形成的漏斗具有两极性 [^1]。 在使用方法上,以网页的ABtest为例,先建立一个有一定流量规模的域,然后对流量进行划分,例如划分为3份用作测试按钮颜色,也可以划分为2份用于不同搜索算法的测试,按钮层和算法层互不影响,可共享使用相同流量 [^4]。 另外,ABtest存在不能二次使用的情况,主要取决于时间周期。随着时间推移会受到多个因素的影响,如用户量、用户质量会变,产品也会迭代,时间周期尽量放在3个月内才可以复用 [^3]。 ```python # 简单示例代码,模拟AB测试流量划分 import random # 模拟总流量 total_traffic = 1000 # 划分流量为两份,用于不同版本测试 version_a_traffic = int(total_traffic * 0.5) version_b_traffic = total_traffic - version_a_traffic # 模拟用户分配到不同版本 users = list(range(total_traffic)) random.shuffle(users) version_a_users = users[:version_a_traffic] version_b_users = users[version_a_traffic:] print(f"Version A 流量: {len(version_a_users)}") print(f"Version B 流量: {len(version_b_users)}") ```
评论
成就一亿技术人!
拼手气红包6.0元
还能输入1000个字符
 
红包 添加红包
表情包 插入表情
 条评论被折叠 查看
添加红包

请填写红包祝福语或标题

红包个数最小为10个

红包金额最低5元

当前余额3.43前往充值 >
需支付:10.00
成就一亿技术人!
领取后你会自动成为博主和红包主的粉丝 规则
hope_wisdom
发出的红包
实付
使用余额支付
点击重新获取
扫码支付
钱包余额 0

抵扣说明:

1.余额是钱包充值的虚拟货币,按照1:1的比例进行支付金额的抵扣。
2.余额无法直接购买下载,可以购买VIP、付费专栏及课程。

余额充值